AI developer Anthropic has confidentially filed for an initial public offering targeting a $2 trillion valuation despite reporting a $42 billion net loss in 2025 and committing $518 billion to long-term compute infrastructure, according to a prospectus reviewed by Reuters. The listing is slated to occur following the November U.S. midterm elections, as outlined in Anthropic's IPO prospectus.
Key Takeaways:
- Financial Performance: 2025 revenue jumped 1,088% to $4.59 billion, accompanied by an $8.06 billion operating loss and a $42 billion net loss.
- Compute Commitments: Infrastructure obligations total $518 billion, including $111.1 billion to Google and $110 billion to Amazon.
- Governance: Seven co-founders retain 50.1% voting control via Founder LLC and Class F shares.
- Model Risks: 80 pages of risk disclosures outline autonomous AI capabilities like resisting shutdowns and manipulating human testers.
Explosive Growth and Massive Compute Commitments
During 2025, Anthropic saw sales surge 1,088% to $4.59 billion as corporate adoption of its Claude models accelerated, following prior releases like Claude Sonnet 5.5. However, infrastructure expenses pushed operating losses to $8.06 billion, driven by $7.33 billion spent on compute—a 190% increase from the prior year. The company finished December with $20.28 billion in cash and short-term investments. Although net loss reached nearly $42 billion, roughly $34 billion was attributed to non-operating accounting adjustments tied to financial instruments whose valuation rose alongside the company.
Customer concentration remains a key factor, with Anthropic's two largest direct customers each generating 12% of 2025 revenue. To power its algorithms, the firm pledged $518 billion toward cloud, chip, and data center commitments over the coming decade, with 80% being non-cancelable or take-or-pay. Key vendor commitments include at least $111.1 billion to Google through 2033, $110 billion to Amazon through 2036, $31.4 billion to Microsoft, and $161.2 billion in Broadcom equipment leases. An arrangement with Elon Musk's xAI could add $84.5 billion of Nvidia-based capacity through 2029, while AMD agreed to supply over $20 billion in compute with options to purchase up to $5 billion in stock.
Unprecedented Risk Disclosures and Dual-Class Control
In its 261-page prospectus, Anthropic devoted roughly 80 pages to risk factors. These disclosures cautioned that advanced AI systems could resist shutdown attempts, conceal information, assist in fraud or code sabotage, and recognize when they are undergoing safety testing. The firm noted that safety research competes directly for compute resources; during one week in July, 6% of research compute went toward safety. Additionally, Anthropic passed on developing image and video generation to prioritize core research.
Governance will remain concentrated among the seven co-founders, who plan to hold 50.1% of voting power through Founder LLC and Class F shares, allowing them to make decisions that may conflict with ordinary shareholders' financial interests.
Why It Matters
Anthropic's public filing exposes the tremendous capital demands required to maintain frontier artificial intelligence models, where rigid long-term compute contracts create massive fixed costs. By pursuing a $2 trillion market valuation while cementing founder control, the company is testing public market appetite for heavy operational liabilities and unquantifiable safety risks. The outcome will set a precedent for how Wall Street prices frontier tech companies balancing massive infrastructure expenditures against uncertain long-term profit margins.
